Tomographic image of melt storage beneath Askja Volcano, Iceland

GEOPHYSICAL RESEARCH LETTERS, VOL. 40, 5040–5046, doi:10.1002/grl.50899, 2013
Tomographic image of melt storage beneath Askja Volcano, Iceland
using local microseismicity
Michael A. Mitchell,1,2 Robert S. White,1 Steve Roecker,3 and Tim Greenfield1
Received 17 June 2013; revised 22 August 2013; accepted 22 August 2013; published 3 October 2013.
[1] We use P wave and S wave arrivals from microseismic earthquakes to construct 3-D tomographic Vp and Vs
images of the magma storage region beneath Askja’s central volcano in the Northern Volcanic Zone of Iceland. A
distinctive ellipsoidal low-velocity anomaly, with both Vp
and Vs velocities 8–12% below the background, is imaged
at 6–11 km depth beneath the caldera. The presence of a
shallow magma chamber is corroborated by geodetic and
gravity studies. The small Vp /Vs anomaly suggests a lack
of pervasive melt. We interpret this anomaly as a region of
multiple sills, some frozen but hot, others containing partial melt. A second, smaller low-velocity anomaly beneath
the main magma storage region may represent a magma
migration pathway. This interpretation is supported by the
close proximity to the anomaly of clusters of deep, magmatically induced earthquakes. However, the location and shape
of this deep anomaly are poorly constrained by the current
data set. Citation: Mitchell, M. A., R. S. White, S. Roecker, and
T. Greenfield (2013), Tomographic image of melt storage beneath
Askja Volcano, Iceland using local microseismicity, Geophys. Res.
Lett., 40, 5040–5046, doi:10.1002/grl.50899.
1. Introduction
[2] Iceland is formed where the mid-Atlantic ridge has
been elevated above sea level as a result of thickened crust
and dynamic support from the underlying mantle plume.
Rifting and volcanism occur within three main neovolcanic
zones (inset, Figure 1). The volcanic zones are subdivided
into local volcanic systems, most of which contain a central
volcano crosscut by a fissure swarm. This study focuses on
the Askja central volcano (sometimes called the Dyngjufjöll
volcanic center) [Sigvaldason, 2002], which lies in the
southern part of the Northern Volcanic Zone (NVZ) and contains a series of three overlapping calderas. Combined, these
calderas cover an area of approximately 45 km2 . Regional
seismic refraction studies, receiver function analysis, and
gravity modeling estimate the crustal thickness beneath
Additional supporting information may be found in the online version
of this article.
1
Department of Earth Sciences, Bullard Laboratories, University of
Cambridge, Cambridge, UK.
2
Now at Department of Earth, Ocean, and Atmospheric Sciences,
Geophysical Inversion Facility (GIF), University of British Columbia,
Vancouver, British Columbia, Canada.
3
Department of Earth and Environmental Sciences, Rensselaer Polytechnic Institute, Troy, New York, USA.
Corresponding author: M. A. Mitchell, Department of Earth, Ocean and
Atmospheric Sciences, Geophysical Inversion Facility (GIF), University of
British Columbia, Earth Sciences Building, 2207 Main Mall, Vancouver,
BC V6T 1Z4, Canada. ([email protected])
©2013. American Geophysical Union. All Rights Reserved.
0094-8276/13/10.1002/grl.50899
the Askja volcanic system to be 30–35 km [Wolfe et al.,
1997; Darbyshire et al., 2000].
[3] The youngest caldera, Öskjuvatn, is partly filled
by a lake and was formed during Askja’s most recent
volcano-tectonic rifting episode in 1875–1876. This episode
consisted of a large explosive, rhyolitic, subPlinianphreatoplinian-Plinian eruption in March 1875, most likely
resulting from an intrusion of basaltic magma into an older,
more silicic reservoir [Sigurdsson and Sparks, 1981]. It
was accompanied by a series of effusive, basaltic eruptions
approximately 60 km to the north in the Sveinagjá graben
from February to November 1875. During the early 20th
century, a series of small basaltic eruptions and one fissure
eruption occurred around the caldera. Öskjuvatn reached its
present size (approximate dimensions: 4.3 km (East-West)
by 3.2 km (North-South) with a maximum depth of 217 m)
in the 1930s following an estimated 2.88 km3 of subsidence
over more than 40 years [Hartley and Thordarson, 2012].
[4] More recent activity includes a small fissure eruption
from the Askja caldera in December 1961 which produced a
9 km long lava flow that breached the northern caldera rim.
Although there has not been any volcanic activity at Askja
since the 1961 eruption, surface expressions of geothermal
activity are still present around Öskjuvatn caldera. Generally, only a few percent, or at most a few tens of percent, of
the melt in a magma chamber is removed during an explosive eruption [Bower and Woods, 1997], so it is likely that
there remains a considerable quantity of melt beneath the
caldera today.
[5] Recent geodetic [Sturkell et al., 2006; Pagli et al.,
2006; Dickinson et al., 2009; Pedersen et al., 2009] and
gravity [de Zeeuw-van Dalfsen et al., 2005] based studies in Askja suggest the presence of a shallow (c. 3 km
deep) magma chamber beneath the caldera. In this study,
we analyze records of microearthquakes recorded by a local
seismometer network to attempt to image the extent of this
magma chamber using seismic arrival time tomography. By
developing a more complete image of Askja’s subcaldera
magma chamber and its associated magmatic plumbing,
researchers can gain important insights into the volcano’s
formation, evolution, and current state of activity.
2. Data Acquisition and Processing
[6] Since July 2006 we have maintained a network of
broadband (0.04–50 Hz) Guralp 6TD three-component seismometers around the Askja caldera (Figure 1). In this study
we analyze arrival times of P and S waves from earthquakes
recorded by this network from 2007 to 2011, supplemented
by observations from the Icelandic Meteorological Office.
Of the 1185 earthquakes used in the inversion, 313 are newly
picked, while the remaining arrival times were picked by
5040
MITCHELL ET AL.: TOMOGRAPHY OF ASKJA VOLCANO, ICELAND
17.2˚W
0
17.0˚W
16.8˚W
16.6˚W
16.4˚W
16.2˚W
16.0˚W
10
(km)
B
Kollóttadyngja
Number of Events
y=−20
0
65.2˚N
Herðubreið
y=−15
20
40
60
80
100 120 140
0
Askja
A
A’
Upptyppingar
y=0
Öskjuvatn
65.0˚N
y=5
Vaðalda
NVZ
y=10
y=15
WV
Z
64.9˚N
y=20
Z
B’
EV
4
6
8
10
12
14
16
18
20
22
24
26
y=25
64.8˚N
x=−25
Earthquake Hypocentral Depths (km)
reið
Dy n
y=−5
Her
ðub
65.1˚N
artö
gjufj
öll y
gl
tri
2
y=−10
x=−20
x=−15
x=−10
x=−5
x=0
x=5
x=10
x=20
x=15
x=25
28
Askja
Herðubreið
Herðubreiðartögl
Vaðalda
Upptyppingar
30
Figure 1. Locations of seismometers (red squares), and earthquakes (colored dots) around Askja used for tomographic
inversion. Earthquakes are color coded by region with Askja, green; HerÄubreiÄ, dark blue; HerÄubreiÄartögl, light blue;
VaÄalda, yellow; and Upptyppingar, red. The olive green shaded regions show the extent of mapped fissure swarms, and the
superimposed grid shows local coordinates used for tomographic inversion. The depth distribution of hypocenters is shown
in histogram to the right of the map. Inset shows regional map of Iceland with small rectangle marking the study region
and fine lines showing the main rift zones: WVZ-Western Volcanic Zone; EVZ-Eastern Volcanic Zone; NVZ-Northern
Volcanic Zone.
Key et al. [2011], Martens et al. [2010], and White et al.
[2011]. Because the resolution and reliability of the final
tomographic model is dependent on the number and orientation of raypaths penetrating each model cell [e.g., Lees,
2007], events were selected to maximize the spatial distribution. Nevertheless, unavoidable gaps in the microseismicity
remain (Figure 1).
[7] Following the selection process, the arrival times were
put through a rigorous manual phase pick refinement process to improve their accuracy. We estimate the average
uncertainties of P and S arrival times to be 0.025 s and
0.05 s, respectively.
3. Velocity Model Inversion
3.1. Regional 1-D Velocity Model
[8] In order to have a reasonable starting model for 3D tomography, we inverted the data set of microseismic
arrival times using VELEST, a damped least squares based
inversion program that can invert for hypocentral locations,
station corrections, and a 1-D velocity model simultaneously
[Kissling et al., 1994]. Due to the interdependent nature
of hypocentral locations, station corrections, and the velocity model, we attempted numerous inversions, with varying
input and control parameters, to produce a stable and geologically reasonable velocity model. The resulting velocity
models are similar to those developed in previous studies,
although Vp and Vs velocities at depths between 16 and
31 km are consistently 2–5% higher than those determined
by Martens et al. [2010] and Key et al. [2011] (Table S2
in the supporting information), which were derived from
two regional seismic refraction surveys in the vicinity:
ICEMELT [Darbyshire et al., 1998] and RRISP [Gebrande
et al., 1980; Menke et al., 1996]. The model developed by
Martens et al. [2010] and Key et al. [2011] will be referred
to as the Askja regional velocity model.
[9] To check the self-consistency of the VELEST derived
1-D velocity model, a series of 1-D reference models were
created using both the VELEST results and the results from
previous studies. After completing an iteration of the 3-D
tomographic inversion using each of these separate starting
models, the resulting data misfit statistics were compared
and the model which minimized the data misfit was selected
as the optimal 1-D velocity model (see Table S1). Figure 2
shows that this optimal 1-D reference model is similar to
Askja’s regional velocity model [Key et al., 2011; Martens
et al., 2010; Darbyshire et al., 1998; Gebrande et al., 1980;
Menke et al., 1996] (see Table S1) with the exception of the
Vp and Vs velocities at depths between 16 and 31 km that are
approximately 3–4% faster. This optimal 1-D velocity model
was used as the starting model for the tomographic inversion
presented in this paper, and the percent changes in velocity
are referenced to this model.
3.2. The 3-D Tomographic Inversion
[10] The 3-D tomographic inversion code used in this
study is essentially the same as that previously used to image
the velocity structure of the San Andreas Fault near Parkfield, California [Roecker et al., 2006] and the western Tien
Shan [Zhiwei et al., 2009]. Travel times and raypaths are
computed using a finite difference eikonal equation solver
based on the methodology of Vidale [1988] as described by
Hole and Zelt [1995]. This approach provides better accuracy in highly heterogeneous environments and decreases
the likelihood of stagnating in a local travel time minima
[Roecker et al., 2006]. P and S velocities are specified at a
5041
MITCHELL ET AL.: TOMOGRAPHY OF ASKJA VOLCANO, ICELAND
[12] In the process of obtaining a final tomographic
model, more than 50 trials were run using different 1-D
reference velocity models, damping parameters, and data
subsets to ensure that possible sources of bias were identified and that the resulting images were consistent. Since
the inversion code has the ability to solve for either Vp and
Vp /Vs or Vp and Vs , trials were performed using both settings.
These trials produced very similar results, thus increasing
our confidence in the large-scale model structures. When
solved for independently, variations in Vs closely track those
of Vp , and only minor variations result from inversions for
Vp /Vs . Because of this similarity, only the Vp model is presented in the main article. The Vs and Vp /Vs models are
shown in Figures S1 and S2. The presented models were produced by solving for Vp and Vp /Vs . Since there are fewer S
wave arrival time picks and they typically have larger uncertainties than the P wave arrival time picks, we expect the Vp
model to have better resolution than the Vs or Vp /Vs models.
4. Results
Figure 2. A comparison of our optimal 1-D reference
velocity model (black lines) with the Askja regional velocity model (blue lines) developed by Martens et al. [2010]
and Key et al. [2011] from two regional seismic refraction
surveys in the vicinity: ICEMELT [Darbyshire et al., 1998]
and RRISP [Gebrande et al., 1980; Menke et al., 1996]. All
depths are in kilometers below sea level with Askja’s caldera
rim reaching its highest point at –1.516 km. As this plot
shows, there is a high degree of correlation between the two
models, although the optimal reference model has slightly
higher Vp and Vs velocities between 16 and 32 km depth.
Since it is difficult to decipher the exact values for depths
and velocities plotted here, the same models are presented in
tabular form within the supporting information (see Tables
S1 and S2).
mesh of nodes and estimated at any point in the model by trilinear interpolation. We chose a grid spacing of 1 km, based
on the anticipated level of resolution of the data set. Model
perturbations are determined by a simultaneous inversion
of velocities and hypocenters using the LSQR algorithm of
Paige and Saunders [1982]. Hypocenters are relocated in the
current model and the procedure iterates until changes in the
variance of the arrival time residuals are insignificant.
[11] To mitigate the introduction of artifacts, a three-cell
wide moving-average window is applied after each iteration to reduce the amplitude of small-scale features. We
chose a damping parameter which allows the data misfit
and model perturbations to decrease steadily at each iteration. Generally, we found that nine iterations were required
before the variance reduction became insignificant as shown
in Figure 3. The final variance is 0.0046 s2 , which is a 41%
reduction from the initial variance but still well above the
variance expected from the uncertainties in the arrival times
(0.0016 s2 ).
[13] The most prominent feature in the tomographic
model (Figures 4 and 5, and S3) is a 8–12% negative velocity
anomaly beneath the Askja caldera. This anomaly extends
approximately 6–11 km in depth and is roughly 6 km wide
in both the E-W and N-S directions (Figure 4). A smaller
low-velocity anomaly extends beneath the main low velocity
body between 13–19 km depth.
[14] We interpret the uniformly lower Vp and Vs in the
main anomaly as evidence of a shallow magma chamber.
However, since only a slight elevation in Vp /Vs is observed
beneath the caldera, and S waves are detected from many
raypaths that cut through the low-velocity region, it is likely
that much of the velocity anomaly is caused by high, but
subsolidus temperature rock with melt distributed in many
small sills rather than in a single large magma chamber.
[15] Based on its shape and location beneath the shallow
magma storage region, and its close proximity to clusters of
Figure 3. A plot of the data misfit variance as a function of
inversion iteration. After approximately 8–10 iterations, the
variance no longer decreases significantly with additional
iterations. Further iterations increase the structural complexity of the model by requiring the addition of small-scale
anomalies which may be artifacts.
5042
MITCHELL ET AL.: TOMOGRAPHY OF ASKJA VOLCANO, ICELAND
deep earthquakes that have been associated with movement
of magma within the lower crust [Soosalu et al., 2010; Key
et al., 2011], we suggest that the smaller and deeper anomaly
represents a magma migration pathway within the crust, possibly as an active/partially crystallized dyke system by which
magma injected into the lower crust is fed into the shallow
magma storage region. At the same time, the size and geometry of this anomaly are not well constrained by the current
data set, so this interpretation is somewhat speculative.
[16] In addition to the low-velocity anomalies, a shallow, 5–10% high velocity anomaly is also imaged directly
beneath the caldera. It is possible that high-density lava
flows within the caldera have produced a near surface, high
velocity core [Brown et al., 1991]. Alternatively, this structure could be a high velocity dome such as the one imaged
beneath Krafla by Brandsdóttir et al. [1997]. Since highvelocity domes are not characteristic of volcanic centers
along the mid-Atlantic ridge, they may be a feature that
develops as part of the interplay between the calderas and
transecting rift zones of Iceland.
[17] We estimate the resolution of the final tomographic
model by determining how well the procedure recovers different synthetic models. Two such models were created: a
standard checker board model (see Figure S4) and a magma
chamber model (see Figure 6). Synthetic phase arrival
time data sets were generated using the same hypocentral
Figure 4. Panels showing x, y, and z sections of the tomographic model (Vp velocities) that slice through the imaged
magma chamber anomaly. (a) A north-south section along x
= –5 km; (b) An east-west section along y = –0.5 km; and (c)
A depth slice from z = 6 km. Figure 4c also shows an outline of the Askja and Öskjuvatn calderas (solid black lines)
and the location of the x and y sections (broken black lines).
Only cells with a ray hit count above 10 are colored.
Figure 5. View of the final tomographic model looking
east on intersecting N-S and E-W sections. The –5.75%
Vp velocity isosurface encloses the imaged magma storage region beneath the caldera. Earthquake hypocenters are
plotted as black dots in this 3-D view. Note the halo of earthquakes that lie above the shallow, subcaldera low-velocity
anomaly. The animation (provided in the supporting information, Figure S3) shows a 360ı rotation of this rendering
to better visualize spatial relationships.
5043
MITCHELL ET AL.: TOMOGRAPHY OF ASKJA VOLCANO, ICELAND
Figure 6. Results of the magma chamber recovery test with three cross sections passing through the inferred magma
chamber beneath the Askja caldera. (a) x, (b) y, and (c) z sections through the true model; (d, e, f ) the corresponding
recovered model sections. As in the results of the checker board recovery test shown in Figure S4, these test results indicate
that the model resolution in the region where the shallow, subcaldera low-velocity anomaly was detected is good enough to
reliably characterize the magma chamber’s location, size, and velocity contrast.
earthquake distribution and station locations as used in the
inversion. These data sets were analyzed in the same manner
as the real data, using the same 1-D starting model, damping parameters, grid cell size, and data/station distribution, to
determine how well the resulting tomographic model could
recover velocity anomalies in different parts of the model.
[18] The synthetic checkerboard model contains a series
of uniformly distributed 4 km by 4 km prisms with an
alternating positive and negative velocity contrast of 5%
with respect to the starting 1-D reference model. Since the
checkerboard model has an even distribution of anomalies
throughout, it is a useful tool for evaluating spatial variations in model resolution. The checkerboard test results (see
Figure S4) show that the data set and procedure are capable
of resolving spatial variations on the order of a few kilometers within the core survey region. In the southern and
western portions of the model, resolution decreases due to a
lack of adequate ray coverage. Since only slight differences
are observed between the recovered Vp and Vs checkerboard
models, the resolution of the Vs model is not significantly
lower than that of the Vp model.
[19] The magma chamber recovery test (Figure 6)
assesses the robustness of the magma chamber image identified in the final tomographic model. The synthetic model
is a single rectangular prism that is similar in size, spatial
location, and velocity contrast (–10% with respect to the 1-D
5044
MITCHELL ET AL.: TOMOGRAPHY OF ASKJA VOLCANO, ICELAND
starting model) to the shallow, subcaldera, low-velocity
anomaly imaged in the shallow magma storage region. As
shown in Figure 6, the maximum amplitude of the magma
chamber anomaly is well recovered, as is the spatial location
of its center. Based on this result, however, it would seem
that the inversion slightly underestimates the size and thickness of the anomaly, putting the center point roughly 0.5 km
shallower than it actually is. Directly beneath the magma
chamber anomaly, there is an apparent pull-down effect that
creates a small tongue of low velocity (velocity contrast of
approximately –3%) material which extends to a depth of
about 12 km, while some slightly positive anomalies (0–2%)
ring the magma chamber in map view. Despite these minor
artifacts, the main features of the anomaly associated with
the magma storage region are well recovered. The lack of a
high velocity anomaly above the magma chamber anomaly
in this test suggests that the shallow high velocity anomaly
imaged beneath the caldera is a real structure and not an
artifact of the inversion.
5. Conclusions
[20] The final tomographic model shows the presence of
a 8–12% low-velocity anomaly beneath the Askja caldera
which extends from 6 to 11 km depth (Figures 4 and 5). Due
to the absence of a large Vp /Vs anomaly, we interpret this
anomaly to be a magma storage region containing high, but
subsolidus temperature rock with melt distributed in many
small sills rather than in a single large magma chamber.
[21] The imaged magma storage region is deeper than the
3 km suggested by GPS and InSAR based subsidence modeling using a point Mogi source [Sturkell et al., 2006; Pagli
et al., 2006], but its center is in a similar position, under
the northern edge of Öskjuvatn. The presence of shallow
seismicity within the caldera extending down to a depth of
4–5 km supports this deeper location of the magma chamber
since microseismicity, which is likely induced by geothermal activity, would not be expected to occur in the region
of magma storage itself but in the cooler more brittle rock
above it. The halo of shallow seismicity above the magma
storage region can be seen in Figure 5 and in the 3-D volume
rotation (Figure S3).
[22] The absence of a deeper low-velocity anomaly
around 16 km depth, where a second magma chamber has
been proposed from Mogi source inversions [Sturkell et al.,
2006; Pagli et al., 2006], supports models which suggest that
geodynamic processes associated with rifting can be used to
account for the regional deformation not accommodated by
a shallow magma chamber [Dickinson et al., 2009; Pedersen
et al., 2009]. Moreover, the inversion result provides some
evidence to support the existence of a melt transport system
extending from the clusters of deep earthquakes [Soosalu
et al., 2010; Key et al., 2011], induced by the injection of
magma into the lower crust, up toward the northern edge of
the magma storage region. However, while this inference is
intriguing, we note that model resolution decreases significantly below 15 km and additional data should be collected
and analyzed to investigate this feature further.
[23] Acknowledgments. Seismometers were borrowed from the Natural Environment Research Council SEIS-UK on loans 857 and 914. Data
from station MKO was kindly made available by the Icelandic Meteorological Office. We thank all those who have assisted in fieldwork and
data analysis since 2006, and especially Janet Key, Hilary Martens, Jon
Tarasewicz, and Heidi Soosalu. Additional thanks go to Janet Key and
Hilary Martens for allowing us to incorporate a large number of their refined
phase arrival picks into this inversion. We would also like to thank Michele
Paulatto and an anonymous reviewer for their insightful and constructive
comments. Figure 1 was produced using Generic Mapping Tools (GMT).
Dept. of Earth Sciences, Cambridge contribution ESC.2858.
[24] The Editor thanks two anonymous reviewers for their assistance in
evaluating this paper.
References
Brandsdóttir, B., W. H. Menke, P. Einarsson, R. S. White, and R. K. Staples
(1997), Faroe-Iceland Ridge Experiment 2. Crustal structure of the Krafla
central volcano, J. Geophys. Res., 102(B4), 7867–7886.
Bower, S. M., and A. W. Woods (1997), Control of magma volatile content and chamber depth on the mass erupted during explosive volcanic
eruptions, J. Geophys. Res., 102, 10,273–10,290.
Brown, G. C., S. P. Everett, H. Rymer, D. W. McGarvie, and I. Foster
(1991), New light on caldera evolution-Askja, Iceland, Geology, 19,
352–355.
Darbyshire, F. A., I. T. Bjarnason, R. S. White, and O. G. Flóvenz
(1998), Crustal structure above the Iceland mantle plume imaged by the
ICEMELT refraction profile, Geophys. J. Int., 135(3), 1131–1149.
Darbyshire, F. A., R. S. White, and K. F. Priestley (2000), Structure of
the crust and uppermost mantle of Iceland from a combined seismic and
gravity study, Earth Sci. Front., 181(3), 409–428.
de Zeeuw-van Dalfsen, E., H. Rymer, F. Sigmundsson, and E. Sturkell
(2005), Net gravity decrease at Askja volcano, Iceland: Constraints on
processes responsible for continuous caldera deflation, 1988–2003, J.
Volcanol. Geotherm. Res., 139, 227–239.
Dickinson, H., T. Masterlark, K. L. Feigl, R. Pedersen, and F. Sigmundsson
(2009), Finite element models for the deformation of the Askja volcanic
complex and rift segment, Iceland, Eos Trans. AGU, 90(52).
Gebrande, H., H. Miller, and P. Einarsson (1980), Seismic structure of
Iceland along the RRISP-Profile, Int. J. Geophys., 47, 239–249.
Hartley, M., and T. Thordarson (2012), Formation of Öskjuvatn caldera at
Askja, North Iceland: Mechanism of caldera collapse and implications
for the lateral flow hypothesis, J. Volcanol. Geotherm. Res., 227–228,
85–101.
Hole, J. A., and B. C. Zelt (1995), 3-D finite-difference reflection traveltimes, Geophys. J. Int., 121(2), 427–434.
Key, J., R. S. White, H. Soosalu, and S. S. Jakobsdóttir (2011), Multiple
melt injection along a spreading segment at Askja, Iceland, Geophys. Res.
Lett., 38(5), 1–5, doi:10.1029/2010GL046264.
Kissling, E., W. L. Ellsworth, D. Eberhart-Phillips, and U. Kradolfer (1994),
Initial reference models in local earthquake tomography, J. Geophys.
Res., 99, 19,635–19,646.
Lees, J. M. (2007), Seismic tomography of magmatic systems, J. Volcanol.
Geotherm. Res., 167(1–4), 37–56.
Martens, H. R., R. S. White, J. Key, J. Drew, H. Soosalu, and S. S.
Jakobsdóttir (2010), Dense seismic network provides new insight into the
2007 Upptyppingar dyke intrusion, Jökull, 60, 47–66.
Menke, W., B. Brandsdóttir, P. Einarsson, and I. T. Bjarnason (1996), Reinterpretation of the RRISP-77 Iceland shear-wave profiles, Geophys. J.
Int., 126(1), 166–172.
Pagli, C., F. Sigmundsson, T. Arnadóttir, P. Einarsson, and E. Sturkell
(2006), Deflation of the Askja volcanic system: Constraints on the deformation source from combined inversion of satellite radar interferograms
and GPS measurements, J. Volcanol. Geotherm. Res., 152(1–2), 97–108.
Paige, C. C., and M. A. Saunders (1982), LSQR: An algorithm for sparse
linear equations and sparse least squares, ACM Trans. Math. Software,
8(1), 43–71.
Pedersen, R., F. Sigmundsson, and T. Masterlark (2009), Rheologic controls on inter-rifting deformation of the Northern Volcanic Zone, Iceland,
Earth Environ. Sci. Trans., 281(1–2), 14–26.
Roecker, S., C. Thurber, K. Roberts, and L. Powell (2006), Refining the
image of the San Andreas Fault near Parkfield, California using a finite
difference travel time computation technique, Tectonophysics, 426(1–2),
189–205.
Sigurdsson, H., and R. S. J. Sparks (1981), Petrology of rhyolitic and mixed
magma ejecta from the 1875 eruption of Askja, Iceland, J. Petrol., 22,
41–84.
Sigvaldason, G. (2002), Volcanic and tectonic processes coinciding with
glaciation and crustal rebound: An early Holocene rhyolitic eruption in
the Dyngjufjöll volcanic centre and the formation of the Askja caldera,
north Iceland, Bull. Volcanol., 64, 192–205.
Soosalu, H., J. Key, R. S. White, C. Knox, P. Einarsson, and S. S.
Jakobsdóttir (2010), Lower-crustal earthquakes caused by magma movement beneath Askja volcano on the north Iceland rift, Bull. Volcanol.,
72(1), 55–62.
Sturkell, E., F. Sigmundsson, and R. Slunga (2006), 1983–2003 decaying rate of deflation at Askja caldera: Pressure decrease in an extensive
5045
MITCHELL ET AL.: TOMOGRAPHY OF ASKJA VOLCANO, ICELAND
magma plumbing system at a spreading plate boundary, Bull. Volcanol.,
68(7–8), 727–735.
Vidale, J. (1988), Finite-difference calculation of travel times, Bull. Seismol.
Soc. Am., 78(6), 2062–2076.
White, R. S., J. Drew, H. R. Martens, J. Key, H. Soosalu, and S. S.
Jakobsdóttir (2011), Dynamics of dyke intrusion in the mid-crust of
Iceland, Earth Planet. Sci. Lett., 304, 300–312.
Wolfe, C., I. T. Bjarnason, J. C. VanDecar, and S. C. Solomon (1997),
Seismic structure of the Iceland mantle plume, Nature, 385, 245–247.
Zhiwei, L., S. Roecker, L. Zhihai, W. Bin, W. Haitao, G. Schelochkov,
and V. Bragin (2009), Tomographic image of the crust and upper mantle
beneath the western Tien Shan from the MANAS broadband deployment: Possible evidence for lithospheric delamination, Tectonophysics,
477(1–2), 49–57.
5046